Skip to main content
Deep Dive is an AI research assistant that reads US municipal disclosure documents and answers credit questions with page-level citations. Research happens inside a session: a multi-turn conversation over a set of documents you choose. You decide what the session is about by adding securities, obligors, individual filings, your own uploads, or any mix of the four. The agent searches those documents, reads the pages that matter, and replies with answers that link back to the pages it relied on.

​
When to use it

  • First read on a credit you’ve never looked at: get an obligor summary, outstanding debt structure, and recent disclosures in a few minutes.
  • Specific question (“What is the debt service coverage on the senior lien?”) that would otherwise mean grepping through hundreds of pages of OS.
  • Multi-year comparison of audited financials or material event notices for the same entity.
  • Side-by-side comparison of two or three obligors in one session. The agent keeps each one’s figures separate and gives each its own row or section.
  • Memo draft: ask the agent to draft a section and edit from there.
It is not a replacement for a credit analyst. It removes the document wrangling so your time goes into judgment.

​
How it works

  1. Click New session and search for the securities, obligors or documents you want to cover, or pick from your own uploads.
  2. The platform pulls the disclosure documents it has for everything you added: official statements, annual disclosures, material event notices, audited financials, and other filings reported to EMMA. Use the document panel to narrow or widen that set.
  3. Ask a free-form question in the chat box.
  4. The agent searches the documents, reads the most relevant pages, and drafts an answer showing which pages it cited.
  5. Ask follow-ups, change the selected documents, add more context, or start a fresh session.
For a security, the session also carries its reference data and its Moody’s rating history, so questions about either are answered without reading a document.

​
Where to enter

A session can be started from several places:
  • The Deep Dive page. Click New session and build the context you want.
  • A security page. Uses that CUSIP.
  • An obligor page. Uses every issue under that obligor.
  • A document in the file viewer. Scopes the session to that document.
  • A Surveillance event. Scopes the session to the documents behind that event.

​
See also

  • Sessions: UI walkthrough, questions, follow-ups, sharing, exports, cancel and retry.
  • Documents: what gets attached, the filter pills, ticking documents in or out.
  • Skills: save instructions you use often and drop them into a question.
  • Writing effective prompts: phrasing patterns for Deep Dive, Tear Sheets, and Data Tables.